MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Given a free group $F$ on $d$ generators and a normal subgroup $H$ of $F$ whose index is finite of prime power order, is there a systematic way to find the numbers of generators of $H/[H,F]$ and of $H/[H,F]H^p$?

share|cite|improve this question
The answer is yes, in the sense that there are algorithms to solve these problems, and it would not be particularly difficult to write programs in a language like GAP or Magma to do so. Is this what you are looking for, or is this more of a theoretical question? – Derek Holt May 9 '13 at 12:15
Thank you Professor Derek Holt, I would like to know if there is an explicit formula (perhaps depending on d and the index of H) to calculate this number of generators. – Yassine Guerboussa May 9 '13 at 12:44
These numbers can certainly depend on the isomorphism type of $F/H$ (and not just on its order). – Derek Holt May 9 '13 at 15:11
up vote 5 down vote accepted

This is really a cohomological question and has a simple cohomological answer. Recall that if a group $G$ acts on an abelian group $M$, then $M_G$ denotes the coinvariants of the action, that is, the quotient of $M$ by the subgroup generated by $\{\text{$m-g(m)$ $|$ $m \in $M, $g \in G$}\}$. The group $F$ acts on $H$ by conjugation, and thus there is an induced action of $F$ on $H_1(H)$.

Key Observation : $H/[H,F] \cong (H_1(H;\mathbb{Z}))_F$ and $H/[H,F]H^p \cong (H_1(H;\mathbb{Z}/p))_F$.

Indeed, we have $H/[H,H] \cong H_1(H;\mathbb{Z})$ and $H/[H,H]H^p \cong H_1(H;\mathbb{Z}/p)$ by definition, and quotienting by $[H,F]$ just kills off the $F$-action.

The other needed ingredient is the 5-term exact sequence in group homology. Given a short exact sequence

$$1 \longrightarrow A \longrightarrow B \longrightarrow C \longrightarrow 1$$

of groups and a ring of coefficients $R$, this 5-term exact sequence takes the form

$$H_2(B;R) \longrightarrow H_2(C;R) \longrightarrow (H_1(A;R))_B \longrightarrow H_1(B;R) \longrightarrow H_1(C;R) \longrightarrow 0.$$

Letting $Q = F/H$, we will apply this to the short exact sequence

$$1 \longrightarrow H \longrightarrow F \longrightarrow Q \longrightarrow 1.$$

The key simplification that occurs is that $H_2(F;R) = 0$ since $F$ is free. We thus get exact sequences

$$0 \longrightarrow H_2(Q;\mathbb{Z}) \longrightarrow H/[F,H] \longrightarrow H_1(F;\mathbb{Z}) \longrightarrow H_1(Q;\mathbb{Z}) \longrightarrow 0$$


$$0 \longrightarrow H_2(Q;\mathbb{Z}/p) \longrightarrow H/[F,H]H^p \longrightarrow H_1(F;\mathbb{Z}/p) \longrightarrow H_1(Q;\mathbb{Z}/p) \longrightarrow 0.$$

If you understand $Q$ enough to calculate its first and second homologies, these short exact sequences allow you to determine $H/[F,H]$ and $H/[F,H]H^p$.

share|cite|improve this answer
Thank you dear professor andy Putman. clearly we can start by a minimally d-generated group Q (I'm interested to the case when Q is a finite p-group), in that case $H_1(Q,Z/p) is isomorphic to the frattini quotient of Q, so it is elementary abelian of rank d. – Yassine Guerboussa May 10 '13 at 10:53
also it is not difficult to prove that $F/[F,F]F^p$ is elementary abelian of rank d, and so is $H_1(F,Z/p)$. It follows from your last exact sequence that $H/[F,H]H^p$ and $H_2(Q,Z/p)$ are isomorphic. So we have only to compute $H_2(Q,Z/p)$. I may ask how much harder to do this? – Yassine Guerboussa May 10 '13 at 11:00
It's nontrivial to compute it, but there is a huge literature on group cohomology, so there are many tools available. To help your search, you should be aware that $H_2(G;\mathbb{Z})$ is also known as the Schur multiplier of $G$. For a particular finite group of reasonable size, by the way, you should be able to compute $H_2$ using GAP. The relevant packages are cohomolo (see and hap (see – Andy Putman May 10 '13 at 16:43
Many thanks dear Andy. – Yassine Guerboussa May 11 '13 at 14:06

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.